Algorithm and implementation of steel head shape recognition based on CRFs

Identifying steel head shapes is an important issue in modern steel rolling production lines. It can be used for plan view control to calculate the length of head cut precisely. With the continuous development of digital image processing technology, edge detection methods are used for object recognition and image analysis. However, in order to adapt to different light conditions, the parameters of edge detectors need to be adjusted frequently. Besides, edge detection methods cannot guarantee the continuity of the segmented region in the steel head shape recognition problem. In this paper, we present an algorithm based on Conditional Random Fields (CRFs) to identify the steel head shape. By using the proposed algorithm and a set of training images, an optimal model is obtained. With this model, new steel images can be processed, and the recognition of steel head shape is achieved. The proposed algorithm can better adapt to the changeable light and need not to alter the parameters frequently. The experimental result show that the algorithm is of strong anti-interference ability and the accuracy is satisfying.